Elastifile's distributed file system designed for use with flash hardware and on-premises and public-cloud storage is adding CloudConnect capabilities for AWS and Google.
Elastifile is out of stealth with a distributed file system its engineers built for use with flash hardware and on-premises and public cloud storage.
The Israeli startup quietly released its Elastifile Cloud File System (ECFS) in November, four years after engineering work began. Next month, Elastifile plans to add new CloudConnect capabilities to enable ECFS to scale out to Amazon and Google public cloud nodes.
ECFS runs on commodity x86 servers equipped with solid-state drives (SSDs). Lightweight virtual controller software installed on each server pools the flash storage and presents it to applications through a global namespace, enabling any server in the cluster to directly access all files.
"A file system is really the best and perhaps the only way to deliver on true hybrid IT without some crazy connectors and a very kluge process," said Andy Fenselau, Elastifile vice president of marketing.
Fenselau said the challenge historically for file systems has been scalability and performance. He said Elastifile's "massive cloud-scale" POSIX-compliant file system could span thousands of on-premises or public-cloud flash-equipped servers and scale out performance linearly with consistent millisecond response time, even in noisy and heterogeneous cloud environments.
Distributed metadata model
Elastifile's key piece of intellectual property is its patented metadata model and Bizur consensus algorithm, Fenselau said. He said other distributed file systems suffer from metadata server bottlenecks. By assuming flash on every node, Elastifile was able to build a distributed metadata model, he said.
"We only run on flash," Fenselau said. "Four years ago, that was a big bet, and we were really nervous about it. Now we're feeling pretty good about that bet. We're seeing all the different tiers of flash."
Each server node serves applications and storage when Elastifile deploys as hyper-converged infrastructure. Customers can also connect application nodes to an Elastifile storage cluster through external NFS, just as a standard NAS system works. ECFS has been tested on and is supported on clusters up to 100 nodes.
Fenselau said customers can deploy Elastifile on Linux-based servers, VMware virtual machines and Amazon Web Services (AWS) Elastic Compute Cloud (EC2) nodes, for use with flash storage in AWS Elastic Block Store (EBS). Elastifile deploys directly into Google Cloud Platform as bare-metal containers, Fenselau said. He said Elastifile plans to add support for Microsoft Azure later this year.
Elastifile includes global deduplication, compression and snapshot shipping (sending only the changes from the previous copy) to help reduce storage capacity needs. The product's CloudConnect feature does the translation to enable customers to tier less active file data to object storage and check it back out at a granular level.
CloudConnect is based on an open format. Users can check out data to other file systems in the cloud or on premises, unlike gateways that use a proprietary format, Fenselau said.
Fenselau said Elastifile has 25 customers, including 16 service providers. He said 13 early adopters use version 1, and the rest are in beta and/or waiting for version 2.0 before going into full production.
Use cases that Elastifile targets include high-performance NAS, mixed workload consolidation in virtualized environments, big data analytics, relational and NoSQL databases, high-performance computing (HPC), and lifting and shifting data and applications to the cloud.
Customer shifting data from Dell EMC, NetApp
Sigmavista IT consulting evaluated Nexenta's ZFS-based storage, Red Hat Ceph Storage, and hyper-converged products from Nutanix and SimpliVity before purchasing Elastifile earlier this year.
The Traun, Austria-based IT service provider currently stores about 10 TB of data in Elastifile and plans to move another 50 TB from its Dell EMC VNX and NetApp FAS systems, according to Daniel Reiter, cloud architect at Sigmavista. Reiter said Sigmavista will use CloudConnect's "lift and shift" capabilities to assist with the NetApp migration.
Reiter said traditional storage is expensive to upgrade and not flexible enough for his business. He said Elastifile gives him flash performance that is easy to manage.
"Eliminating caching, using all-flash, means we get 100% consistent performance, which is what our customers need," Reiter said.
Sigmavista uses Elastifile in its own data center but eventually plans to scale out its deployment to public clouds.
Veteran leadership team
Reiter was impressed by the experience of Elastifile's leadership team. Elastifile founder and CTO Shahar Frank also started all-flash array vendor XtremIO, led development of the storage stack for virtualization at Red Hat, and served as chief architect at Exanet. EMC acquired XtremIO and Dell acquired Exanet, both long before the two IT giants merged last year.
Other Elastifile executives have extensive experience in storage, flash, virtualization and application optimization technologies through their work with vendors such as IBM (XIV), Red Hat, SanDisk, and Anobit (acquired by Apple).
Since its 2013 founding, Elastifile has raised more than $50 million from venture capitalists and technology vendors such as Battery Ventures, Lightspeed Venture Partners, Cisco and Western Digital. The startup has offices in Herzliya, Israel, and Santa Clara, Calif.
Pricing for Elastifile's software is consumption- and subscription-based, at a starting list cost of two cents per GB per month. Fenselau said Elastifile would offer two packages: one for on-premises deployments and another for cross-cloud deployments and cloud mobility capabilities such as CloudConnect. Customers need to supply their own server and flash hardware.
"It's really interesting to see a company that's delivering a software-only model make such a strong recommendation to be on a flash-based node," said George Crump, president of Storage Switzerland. "Usually when you make that recommendation, you're also then supplying the node."
Crump said large enterprises and service provider customers might prefer the "bring your own hardware" approach. "But as they evolve, just like probably everybody else, they'll need to provide the hardware or have a close relationship with a Lenovo or Supermicro" to provide more of a turnkey option to enable customers to "just plug it in and go," Crump said.
Crump said Elastifile is among a group of startups that focus both on moving data to accelerate it and shifting data to cheaper storage. Others include Primary Data and ioFABRIC. He said historically vendors either focused on accelerating data or moving it to cheaper storage.
Henry Baltazar, a storage research director at 451 Research, said the software-defined storage (SDS) competition from startups includes Datera, FalconStor, Formation Data Systems, Hedvig, Nexenta, Qumulo, Quobyte and Zadara. Larger vendors, such as Dell EMC, IBM and NetApp, have also entered the SDS space, as have VMware, with vSAN, and Red Hat and SUSE, with Ceph.
"Elastifile feels its advanced management capabilities, such as storage [quality of service] QoS and flash native architecture, can be differentiators" in a crowded market, Baltazar wrote in an email.
Steve Duplessie, founder and senior analyst at Enterprise Strategy Group, said Elastifile's true competition is the "old way of doing things." He said nearly all corporate file shares currently run on either NetApp systems or Windows Server shares. The Elastile file system would enable users to have a single virtual file server that everyone could access from any physical location.
"I don't need to have servers all over the globe anymore," Duplessie wrote in an email.